Global Code and Functions

Run this first to import modules and global functions

#Modules
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
import json
from IPython.display import Image, display
import re
import math
import torch
import os
import cv2
from collections import Counter
import pandas as pd

# Models
from scipy.spatial import distance
from sklearn.neighbors import NearestNeighbors
from sklearn.manifold import TSNE
from sklearn.cluster import KMeans
from sklearn.decomposition import PCA
from sklearn.metrics import silhouette_score
from transformers import CLIPProcessor, CLIPModel
from sklearn.metrics.pairwise import cosine_similarity




# Global Functions

def get_top50_ann(target_embedding, embeddings):
    nn = NearestNeighbors(n_neighbors=51, metric='cosine', algorithm='brute')
    nn.fit(embeddings)
    distances, indices = nn.kneighbors([target_embedding])
    return indices[0][1:]  # Skip the first index because it's the target itself

def get_top50_euclidean(target_embedding, embeddings):
    distances = [distance.euclidean(target_embedding, emb) for emb in embeddings]
    indices = np.argsort(distances)[1:51]  # Skip the first index because it's the target itself
    return indices

def display_image(index):
    display(Image(filename=f"thumbnails_folder2large/{g_movie_embeddings[index]['input']}"))

def display_images(indices, embeddings):
    fig, axes = plt.subplots(10, 10, figsize=(20, 10))
    for i, ax in enumerate(axes.flat):
        ax.imshow(plt.imread(f"thumbnails_folder2large/{g_movie_embeddings[indices[i]]['input']}"))
        ax.axis('off')
    plt.show()

def display_images_first_x_last_x(indices, first_x, last_x, cluster_n=0):
    # Select the first_x and last_x indices
    selected_indices = indices[:first_x] + indices[-last_x:]
    
    # Calculate the number of rows and columns for the subplot
    total_images = first_x + last_x
    cols = 10
    rows = math.ceil(total_images / cols)
    
    fig, axes = plt.subplots(rows, cols, figsize=(20, 2 * rows))
    axes = axes.ravel()  # Flatten the axes array
    
    # Hide all axes
    for ax in axes:
        ax.axis('off')
    
    # Display images on the first len(selected_indices) axes
    for i, idx in enumerate(selected_indices):
        axes[i].imshow(plt.imread(f"thumbnails_folder2large/{g_movie_embeddings[idx]['input']}"))
        axes[i].axis('on')
    
    plt.tight_layout()
    plt.title(f"Cluster {cluster_n} -  First {first_x} and Last {last_x} Images - Total Images in Cluster: {len(indices)}")
    plt.show()

def display_cluster_images(cluster_labels, cluster_number):
    # Get indices of images in the cluster
    indices = [i for i, label in enumerate(cluster_labels) if label == cluster_number]
    
    # Display images
    display_images(indices)

def display_cluster_images_first_last_x(cluster_labels, cluster_number, first_x, last_x):
    # Get indices of images in the cluster
    indices = [i for i, label in enumerate(cluster_labels) if label == cluster_number]
    
    # Display images
    display_images_first_x_last_x(indices, first_x, last_x, cluster_number)

def find_and_remove_intro_and_subtitles(g_only_embeddings, threshold=0.7):
    # Load the CLIP model and processor
    model = CLIPModel.from_pretrained("openai/clip-vit-large-patch14")
    processor = CLIPProcessor.from_pretrained("openai/clip-vit-large-patch14")

    # Get the text embeddings for the intro and subtitles
    inputs = processor(text=["Image of Walt Disney Movie Intro", "Walt Disney Movie Intro", "Movie closing credits", "Movie end credits", "Image with lots of closing credits"], return_tensors="pt", padding=True)
    text_embeddings = model.get_text_features(**inputs)
    text_embeddings_np = text_embeddings.detach().numpy()

    # Calculate the cosine similarity between the text embeddings and the movie embeddings
    similarities = cosine_similarity(text_embeddings_np, g_only_embeddings)

    # Find the indices of the embeddings that are similar to the intro and subtitles
    intro_subtitle_indices = np.where(similarities.max(axis=0) > threshold)[0]

    #print("Number of images(Intro and Closing credits) to remove:", intro_subtitle_indices)

    # Create new lists that exclude the intro and subtitles
    new_g_movie_embeddings = [emb for i, emb in enumerate(g_movie_embeddings) if i not in intro_subtitle_indices]
    new_g_only_embeddings = np.array([emb for i, emb in enumerate(g_only_embeddings) if i not in intro_subtitle_indices])

    return new_g_movie_embeddings, new_g_only_embeddings

# Global Variables
g_movie_embeddings = json.load(open("honey_i_shrunk_the_kids_movie_embeddings_1_second.json"))
g_only_embeddings = np.array([emb['embedding'] for emb in g_movie_embeddings])

g_movie_embeddings, g_only_embeddings = find_and_remove_intro_and_subtitles(g_only_embeddings, threshold=0.237)

Hypothesis 1

dog_idx = 5038
mower_idx = 4733
ant_idx = 3230

model = CLIPModel.from_pretrained("openai/clip-vit-large-patch14")
processor = CLIPProcessor.from_pretrained("openai/clip-vit-large-patch14")

with open("honey_i_shrunk_the_kids_movie_embeddings_1_second.json", 'r') as file:
    movie_embeddings = json.load(file)

def euclidean_distance(array1, array2):
    # Convert the arrays to NumPy arrays
    array1_np = np.array(array1)
    array2_np = np.array(array2)
    
    # Calculate the Euclidean distance
    distance = np.linalg.norm(array1_np - array2_np)
    return distance

def find_and_display_matches(text_queries, top_k=5):
    inputs = processor(text=text_queries, return_tensors="pt", padding=True)
    text_embeddings = model.get_text_features(**inputs)
    text_embeddings_np = text_embeddings.detach().numpy()
    movie_embeddings_np = np.array([movie['embedding'] for movie in movie_embeddings])
    similarities = cosine_similarity(text_embeddings_np, movie_embeddings_np)
    for index, text_query in enumerate(text_queries):
        print(f"Top matches for: {text_query}")
        top_indices = np.argsort(similarities[index])[::-1][:top_k]
        for i in top_indices:
            frame = movie_embeddings[i]['input']
            print(f"Displaying frame: {frame}")
            display(Image(filename=f'thumbnails_folder2large/{frame}'))

def plot_euclidean_distance_from(target_idx):
    target = movie_embeddings[target_idx]
    image_path = image_path = 'thumbnails_folder2large/' + target["input"]
    # Display the image
    display(Image(filename=image_path))

    index_to_distance = []

    # Iterate through the input list
    for emb in movie_embeddings:
        current_dist = euclidean_distance(emb["embedding"], target["embedding"])
        index_to_distance.append(current_dist)

    # Create a plot using Seaborn
    sns.set(style="whitegrid")  # Set the style
    plt.figure(figsize=(10, 6))  # Set the figure size
    sns.lineplot(x=range(len(index_to_distance)), y=index_to_distance)  # Plot the array with index as x-axis
    plt.xlabel("Index")  # Set the x-axis label
    plt.ylabel("Distance")  # Set the y-axis label
    plt.title("Distance from Target Over Film")  # Set the title
    plt.show()  # Show the plot

def display_surrounding_frames(target_idx, frame_range=5):
    idx_begin = target_idx - frame_range
    idx_end = target_idx + frame_range + 1

    display_frames = movie_embeddings[idx_begin:idx_end]
    for emb in display_frames:
        print(f'frame {emb["input"]}')
        image_path = 'thumbnails_folder2large/' + emb["input"]
        display(Image(filename=image_path))

Exploring the Dog

Dislpay the images

display_surrounding_frames(dog_idx)

frame thumbnail_5034.jpg

frame thumbnail_5035.jpg

frame thumbnail_5036.jpg

frame thumbnail_5037.jpg

frame thumbnail_5038.jpg

frame thumbnail_5039.jpg

frame thumbnail_5040.jpg

frame thumbnail_5041.jpg

frame thumbnail_5042.jpg

frame thumbnail_5043.jpg

frame thumbnail_5044.jpg

plot_euclidean_distance_from(dog_idx)

# Using the following to explore 500, 2100 and 3400
#display_surrounding_frames(3400, frame_range=40)
# Discovered the dog in
# thumbnail_0514.jpg
# thumbnail_2078.jpg
# thumbnail_2131.jpg
# thumbnail_3427.jpg
# Indexes are are 0 based
display_surrounding_frames(513, frame_range=2)

frame thumbnail_0512.jpg

frame thumbnail_0513.jpg

frame thumbnail_0514.jpg

frame thumbnail_0515.jpg

frame thumbnail_0516.jpg

The dog is watching the father's scientific experiment.

display_surrounding_frames(2077, frame_range=2)

frame thumbnail_2076.jpg

frame thumbnail_2077.jpg

frame thumbnail_2078.jpg

frame thumbnail_2079.jpg

frame thumbnail_2080.jpg

The dog seems to notice something outside.

display_surrounding_frames(2130, frame_range=2)

frame thumbnail_2129.jpg

frame thumbnail_2130.jpg

frame thumbnail_2131.jpg

frame thumbnail_2132.jpg

frame thumbnail_2133.jpg

The parents are distracted, and the dog wants to investigate what is going on outside, once again exhibiting a higher sense of awareness than the humans.

display_surrounding_frames(3426, frame_range=2)

frame thumbnail_3425.jpg

frame thumbnail_3426.jpg

frame thumbnail_3427.jpg

frame thumbnail_3428.jpg

frame thumbnail_3429.jpg

Here the dog is spinning the father around and disrupting his search for the children, which is an example of a comedic scene.

Exploring the Lawnmower

Dislpay the images

display_surrounding_frames(mower_idx)

frame thumbnail_4729.jpg

frame thumbnail_4730.jpg

frame thumbnail_4731.jpg

frame thumbnail_4732.jpg

frame thumbnail_4733.jpg

frame thumbnail_4734.jpg

frame thumbnail_4735.jpg

frame thumbnail_4736.jpg

frame thumbnail_4737.jpg

frame thumbnail_4738.jpg

frame thumbnail_4739.jpg

plot_euclidean_distance_from(mower_idx)

# Display around 790
#display_surrounding_frames(790, frame_range=40)
display_surrounding_frames(790, frame_range=5)

frame thumbnail_0786.jpg

frame thumbnail_0787.jpg

frame thumbnail_0788.jpg

frame thumbnail_0789.jpg

frame thumbnail_0790.jpg

frame thumbnail_0791.jpg

frame thumbnail_0792.jpg

frame thumbnail_0793.jpg

frame thumbnail_0794.jpg

frame thumbnail_0795.jpg

frame thumbnail_0796.jpg

Exploring the Ant

Dislpay the images

display_surrounding_frames(ant_idx)

frame thumbnail_3226.jpg

frame thumbnail_3227.jpg

frame thumbnail_3228.jpg

frame thumbnail_3229.jpg

frame thumbnail_3230.jpg

frame thumbnail_3231.jpg

frame thumbnail_3232.jpg

frame thumbnail_3233.jpg

frame thumbnail_3234.jpg

frame thumbnail_3235.jpg

frame thumbnail_3236.jpg

plot_euclidean_distance_from(ant_idx)

# Display around 2180
#display_surrounding_frames(2180, frame_range=40)
display_surrounding_frames(2180, frame_range=1)

frame thumbnail_2180.jpg

frame thumbnail_2181.jpg

frame thumbnail_2182.jpg

# Display around 4400
#display_surrounding_frames(4400, frame_range=40)
display_surrounding_frames(4435, frame_range=1)

frame thumbnail_4435.jpg

frame thumbnail_4436.jpg

frame thumbnail_4437.jpg

The scene around 4435 shows an ant fighting a scorpion and appearing to save the kids.

Use CLIP to ask about 'a photo of an ant fighting a scorpion' to see what it returns.

exploration_query = ['a photo of an ant fighting a scorpion'] 
find_and_display_matches(exploration_query, top_k=5)

Top matches for: a photo of an ant fighting a scorpion
Displaying frame: thumbnail_4433.jpg

Displaying frame: thumbnail_3469.jpg

Displaying frame: thumbnail_4132.jpg

Displaying frame: thumbnail_4133.jpg

Displaying frame: thumbnail_3292.jpg

Hypothesis 2

- Scene Consistency and Transition with dimensionality reduction and clustering -

Cluster Analysis

We will compare clustering with t-SNE (t-Distributed Stochastic Neighbor Embedding) and PCA (Principal Component Analysis) dimensionality reduction algorithms.

t-SNE

# Using t-SNE to embed the vectors into 2D
tsne = TSNE(n_components=2, random_state=42)
tSNE_embedded_vectors = tsne.fit_transform(g_only_embeddings)

PCA

# Using PCA to embed the vectors into 2D
pca = PCA(n_components=2)
PCA_embedded_vectors = pca.fit_transform(g_only_embeddings)

Cluster t-SNE and PCA with K-Means and display Silhoutte Score

Silhouette Score: Measures how similar an object is to its own cluster compared to other clusters. The score ranges from -1 to 1, where a high value indicates that the object is well matched to its own cluster and poorly matched to neighboring clusters.

Lets find the best clustering number for t-SNE and PCA...

X = tSNE_embedded_vectors

# Range of clusters to try
num_clusters = range(2, 20)

# List to hold silhouette scores
sil_scores = []

# Loop over number of clusters
for k in num_clusters:
    # Perform clustering
    kmeans = KMeans(n_init="auto", n_clusters=k, random_state=42).fit(X)
    
    # Get cluster labels
    labels = kmeans.labels_
    
    # Compute silhouette score and append to list
    sil_score = silhouette_score(X, labels)
    sil_scores.append(sil_score)

# Plot silhouette scores
plt.plot(num_clusters, sil_scores, 'bx-')
plt.title('t-SNE')
plt.xlabel('k (number of clusters)')
plt.ylabel('Silhouette Score')
plt.show()



X = PCA_embedded_vectors

# Range of clusters to try
num_clusters = range(2, 20)

# List to hold silhouette scores
sil_scores = []

# Loop over number of clusters
for k in num_clusters:
    # Perform clustering
    kmeans = KMeans(n_init="auto", n_clusters=k, random_state=42).fit(X)
    
    # Get cluster labels
    labels = kmeans.labels_
    
    # Compute silhouette score and append to list
    sil_score = silhouette_score(X, labels)
    sil_scores.append(sil_score)

# Plot silhouette scores
plt.plot(num_clusters, sil_scores, 'bx-')
plt.title('PCA')
plt.xlabel('k (number of clusters)')
plt.ylabel('Silhouette Score')
plt.show()

Lets cluster with t-SNE and PCA best Silhoutte Scores.

# Performing KMeans clustering with best k silhoutte score.
kmeans = KMeans(n_init="auto", n_clusters=19, random_state=42)
tSNE_clusters = kmeans.fit_predict(tSNE_embedded_vectors)

kmeans = KMeans(n_init="auto", n_clusters=4, random_state=42)
PCA_clusters = kmeans.fit_predict(PCA_embedded_vectors)


# Extracting numbers from file names for labels
labels = [re.search(r'\d+', vector['input']).group() for vector in g_movie_embeddings]

#t-SNE
# Plotting the embedded vectors with cluster coloring
sns.set_theme()
plt.figure(figsize=(12, 8))  # Adjust the figure size as needed
sns.scatterplot(x=tSNE_embedded_vectors[:, 0], y=tSNE_embedded_vectors[:, 1], hue=tSNE_clusters, palette='bright', legend='full', s=100)
for i, vec in enumerate(tSNE_embedded_vectors):
    plt.text(vec[0] + 0.02, vec[1] + 0.02, labels[i], fontsize=6)  # Adding labels
plt.xlabel('Dimension 1')
plt.ylabel('Dimension 2')
plt.title('t-SNE Embedded Vectors with KMeans Clustering (k=16)')
plt.legend(title='Cluster')
plt.show()

#PCA
# Plotting the embedded vectors with cluster coloring
sns.set_theme()
plt.figure(figsize=(12, 8))  # Adjust the figure size as needed
sns.scatterplot(x=PCA_embedded_vectors[:, 0], y=PCA_embedded_vectors[:, 1], hue=PCA_clusters, palette='bright', legend='full', s=100)
for i, vec in enumerate(PCA_embedded_vectors):
    plt.text(vec[0] + 0.02, vec[1] + 0.02, labels[i], fontsize=6)  # Adding labels
plt.xlabel('Dimension 1')
plt.ylabel('Dimension 2')
plt.title('PCA Embedded Vectors with KMeans Clustering (k=4)')
plt.legend(title='Cluster')
plt.show()

Let sample images in t-SNE clusters

unique_clusters = set(tSNE_clusters)

for cluster in unique_clusters:
    display_cluster_images_first_last_x(tSNE_clusters, cluster, 10, 10)

Lets sample images in PCA clusters

unique_clusters = set(PCA_clusters)

for cluster in unique_clusters:
    display_cluster_images_first_last_x(PCA_clusters, cluster, 10, 10)

t-SNE and PCA with K-means clusters over a timeline

By looking at these plots, we can see if frames from the same cluster tend to occur close together in time, which might indicate that the clustering is capturing some meaningful structure in the movie. For example, all the frames from a particular scene might be grouped into the same cluster.

# Convert labels to timestamps by dividing by frame rate
# Assuming `frame_rate` is the frame rate of the movie
frame_rate = 24
timestamps = [int(label) / frame_rate for label in labels]

# Create a timeline plot for the t-SNE clusters
plt.figure(figsize=(12, 6))
plt.scatter(timestamps, tSNE_clusters, c=tSNE_clusters, cmap='viridis')
plt.xlabel('Time')
plt.ylabel('Cluster')
plt.title('t-SNE Clusters Over Time')
plt.colorbar(label='Cluster')
plt.show()

# Create a timeline plot for the PCA clusters
plt.figure(figsize=(12, 6))
plt.scatter(timestamps, PCA_clusters, c=PCA_clusters, cmap='viridis')
plt.xlabel('Time')
plt.ylabel('Cluster')
plt.title('PCA Clusters Over Time')
plt.colorbar(label='Cluster')
plt.show()

Hypothesis 3

Find a model that can identify objects from movie frame images

Utilizing the Yolo model for recognizing and counting objects

import torch
import os
from PIL import Image
import matplotlib.pyplot as plt
import cv2
from collections import Counter


model = torch.hub.load("ultralytics/yolov5", "yolov5s", pretrained=True)


image_directory = "thumbnails_folder2large/"

# List the first 100 images in the image file directory
image_files = [
    f
    for f in os.listdir(image_directory)
    if os.path.isfile(os.path.join(image_directory, f))
]
image_files = image_files[:100]

# Initialize a Counter to keep track of object counts
total_counts = Counter()


for image_file in image_files:

    image_path = os.path.join(image_directory, image_file)

    results = model(image_path)

    print(f"Results for {image_file}:")
    df = results.pandas().xyxy[0]  # Results as DataFrame
    print(df)

    # Update the total counts of objects
    counts = df["name"].value_counts()
    total_counts.update(counts)

    results.show()
    results.save(save_dir="output/")

    object_counts = df["name"].value_counts()
    print("Object counts:", object_counts)
    print("\n")

Using cache found in /home/stace/.cache/torch/hub/ultralytics_yolov5_master
YOLOv5 🚀 2024-4-27 Python-3.10.12 torch-2.3.0+cu121 CPU

Fusing layers... 
YOLOv5s summary: 213 layers, 7225885 parameters, 0 gradients, 16.4 GFLOPs
Adding AutoShape... 

Results for thumbnail_0795.jpg:
         xmin       ymin        xmax        ymax  confidence  class  \
0  165.206284   6.893837  215.419052  170.662994    0.905672      0   
1  101.212090  35.575703  155.678467  141.318115    0.815038      0   
2  100.350342   0.000000  129.249664   26.236727    0.532667     58   
3   94.354683  39.541080  110.575081   63.907471    0.360822     58   
4  126.063278  13.044321  137.669662   28.871672    0.265851     58   
5   86.855400   0.000000  116.105843   25.523857    0.260697     58   

           name  
0        person  
1        person  
2  potted plant  
3  potted plant  
4  potted plant  
5  potted plant  

Saved 1 image to output

Object counts: name
potted plant    4
person          2
Name: count, dtype: int64


Results for thumbnail_2778.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output2

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_2599.jpg:
         xmin       ymin        xmax        ymax  confidence  class      name
0   39.705357   3.163864  143.259903  170.139664    0.871375      0    person
1  148.172928  66.428223  243.559738  170.312714    0.830257      0    person
2  228.076645  61.408001  319.787598   99.793732    0.372595     25  umbrella

Saved 1 image to output3

Object counts: name
person      2
umbrella    1
Name: count, dtype: int64


Results for thumbnail_5102.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   62.999626   0.411362  243.993866  169.815109    0.888487      0  person
1  201.136368   2.532188  319.873016  170.904419    0.368096      0  person
2  201.111191  59.719219  245.755753  106.706238    0.308551      0  person

Saved 1 image to output4

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_2869.jpg:
        xmin      ymin        xmax        ymax  confidence  class    name
0  44.419312  2.505157  279.158417  167.806213    0.842688      0  person

Saved 1 image to output5

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4763.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output6

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_1841.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output7

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_4344.jpg:
        xmin      ymin        xmax        ymax  confidence  class    name
0  69.193626  2.837288  314.933716  171.205292    0.704568      0  person
1   2.009521  1.623985  135.926559  170.301331    0.579770      0  person

Saved 1 image to output8

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_2830.jpg:
   xmin      ymin        xmax        ymax  confidence  class    name
0   0.0  2.524231  140.201889  168.760849    0.432279      0  person

Saved 1 image to output9

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_0764.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output10

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_4769.jpg:
         xmin       ymin        xmax       ymax  confidence  class  name
0  200.240234  44.231087  240.104523  86.270424    0.276669     21  bear

Saved 1 image to output11

Object counts: name
bear    1
Name: count, dtype: int64


Results for thumbnail_5230.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  111.571434  53.508400  234.566559  150.327423    0.877162      0  person
1  193.450363   5.458885  289.558289  146.941132    0.750499      0  person

Saved 1 image to output12

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_4402.jpg:
        xmin      ymin       xmax        ymax  confidence  class  name
0  58.711868  66.00354  317.49469  168.783325    0.405987     55  cake

Saved 1 image to output13

Object counts: name
cake    1
Name: count, dtype: int64


Results for thumbnail_0838.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0   10.887814    9.057289  121.294357  170.662140    0.918634      0   
1  132.176971   72.398933  225.301208  171.744598    0.889513      0   
2  214.722260  101.096077  230.285492  122.825005    0.479683     67   
3  262.358063   95.709274  319.863617  172.149094    0.373231      1   
4  134.295975   75.295395  146.226730  101.754227    0.364346     58   
5  140.245438   44.690025  156.069687   70.838554    0.332648     58   
6  127.589920   87.786789  137.921951  104.485153    0.320908     58   
7  263.842987   95.603409  320.000000  171.102280    0.270932     58   
8  126.522469  107.228600  148.275803  143.455444    0.250760     58   

           name  
0        person  
1        person  
2    cell phone  
3       bicycle  
4  potted plant  
5  potted plant  
6  potted plant  
7  potted plant  
8  potted plant  

Saved 1 image to output14

Object counts: name
potted plant    5
person          2
cell phone      1
bicycle         1
Name: count, dtype: int64


Results for thumbnail_3214.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  144.634552  32.005585  317.515961  171.461853    0.928847      0  person
1   20.182396  51.222668  173.023438  170.957565    0.680408      0  person

Saved 1 image to output15

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_3277.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0    0.053280  49.604023   75.099426  169.629684    0.827735      0  person
1   73.538979  49.266884  160.660828  168.734863    0.820189      0  person
2  154.055847  78.233902  201.372101  170.313568    0.729479      0  person

Saved 1 image to output16

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_1786.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  73.028404  55.309113  160.393433  170.868866    0.653283      0  person

Saved 1 image to output17

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4734.jpg:
         xmin       ymin        xmax        ymax  confidence  class   name
0  280.429657  53.827934  319.387726   97.772301    0.737459     56  chair
1  114.833496  40.400749  223.138916  133.413239    0.473742      7  truck
2  190.397873  49.629734  223.321640   63.169189    0.339173      2    car

Saved 1 image to output18

Object counts: name
chair    1
truck    1
car      1
Name: count, dtype: int64


Results for thumbnail_1414.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output19

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_0402.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  187.116348  98.901291  218.031570  143.041443    0.862557      0  person
1  214.157837  83.272797  240.787567  142.929199    0.589160      0  person

Saved 1 image to output20

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_0604.jpg:
   xmin  ymin        xmax        ymax  confidence  class    name
0   0.0   0.0  267.941467  168.336533    0.653137      0  person

Saved 1 image to output21

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_1578.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  179.073669  27.716438  230.254181  168.477631    0.803939      0  person

Saved 1 image to output22

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3638.jpg:
         xmin      ymin        xmax        ymax  confidence  class    name
0    0.524506  2.683617  190.530685  170.092987    0.890239      0  person
1  177.725220  3.032242  318.074402  170.370590    0.403448     15     cat

Saved 1 image to output23

Object counts: name
person    1
cat       1
Name: count, dtype: int64


Results for thumbnail_2130.jpg:
        xmin       ymin        xmax        ymax  confidence  class name
0  66.485329  15.057289  202.649689  163.535858    0.825101     15  cat

Saved 1 image to output24

Object counts: name
cat    1
Name: count, dtype: int64


Results for thumbnail_0904.jpg:
         xmin      ymin        xmax        ymax  confidence  class    name
0  143.296783  2.758759  296.191437  169.471558    0.869524      0  person

Saved 1 image to output25

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_1196.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   40.405964  90.664719   84.403656  171.545609    0.858486      0  person
1  183.029037   3.579781  319.300720  171.077515    0.574772      0  person
2  138.948303  39.238800  263.657074  170.145706    0.465853      0  person

Saved 1 image to output26

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_4256.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0  146.819992    9.440872  320.000000  172.989639    0.945141      0  person
1   20.983765   13.298370  161.444916  170.617905    0.918598      0  person
2   86.343025  113.282387  101.466789  170.770935    0.297957     27     tie

Saved 1 image to output27

Object counts: name
person    2
tie       1
Name: count, dtype: int64


Results for thumbnail_4935.jpg:
        xmin       ymin        xmax       ymax  confidence  class    name
0  88.580727  26.743401  248.351013  170.39566    0.623986      0  person

Saved 1 image to output28

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_2105.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output29

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_0214.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  176.159821    8.380219  287.971924  141.245422    0.850449      0   
1  294.993713  102.353477  319.953003  145.961212    0.844055     41   
2   11.467091  104.170700  123.098709  171.487625    0.587605     15   
3    1.286644   72.242310  112.192795  144.937378    0.459279     58   
4  258.418427   70.611099  308.974274  114.212433    0.433455     56   
5  130.409546  128.525208  142.579285  144.590881    0.351891     41   
6   96.898674   80.701042  166.870911  137.720917    0.331840     13   
7  266.991577  120.622444  297.645996  145.624557    0.289562     41   
8  166.048965   52.058750  179.461533   83.111313    0.265729     39   

           name  
0        person  
1           cup  
2           cat  
3  potted plant  
4         chair  
5           cup  
6         bench  
7           cup  
8        bottle  

Saved 1 image to output30

Object counts: name
cup             3
person          1
cat             1
potted plant    1
chair           1
bench           1
bottle          1
Name: count, dtype: int64


Results for thumbnail_1313.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0  130.926376    2.710327  318.124573  170.938354    0.923482      0  person
1  205.433197  129.331558  235.381714  173.000000    0.863757     27     tie

Saved 1 image to output31

Object counts: name
person    1
tie       1
Name: count, dtype: int64


Results for thumbnail_4161.jpg:
        xmin      ymin        xmax        ymax  confidence  class    name
0  95.329269  1.505486  309.440125  170.497589    0.743999      0  person

Saved 1 image to output32

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_1218.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   59.464005   5.452637  199.835617  171.995453    0.924212      0  person
1  166.318619  47.425896  275.354126  171.355835    0.819016      0  person

Saved 1 image to output33

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_3185.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  68.541718  22.858665  100.686615  115.881477    0.650451      0  person

Saved 1 image to output34

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3004.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   16.467949   7.679947  185.351685  171.601318    0.788966      0  person
1  177.684647  13.900711  318.367249  171.006119    0.769397      0  person

Saved 1 image to output35

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_2117.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output36

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_2194.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  128.264832   17.223923  246.417358  157.115829    0.512254      0   
1  221.005920  111.913651  320.000000  171.219070    0.356592     77   
2   56.877853   40.796303  114.606705   88.694984    0.286311     50   

         name  
0      person  
1  teddy bear  
2    broccoli  

Saved 1 image to output37

Object counts: name
person        1
teddy bear    1
broccoli      1
Name: count, dtype: int64


Results for thumbnail_2005.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output38

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_3695.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   23.640350  66.985527  170.474426  166.495819    0.357245     16     dog
1  133.369095  29.806076  287.624939  123.378754    0.354583      0  person

Saved 1 image to output39

Object counts: name
dog       1
person    1
Name: count, dtype: int64


Results for thumbnail_3129.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0  251.405396   58.205311  320.000000  171.264816    0.925341      0  person
1  215.099197  129.501480  260.699585  172.702438    0.904707      0  person

Saved 1 image to output40

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_2384.jpg:
        xmin      ymin        xmax        ymax  confidence  class    name
0  87.236252  6.909767  196.455521  171.686859    0.938032      0  person

Saved 1 image to output41

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_1120.jpg:
         xmin       ymin        xmax        ymax  confidence  class      name
0  195.474915  33.009022  239.872314  151.447754    0.705667      0    person
1  171.965424  74.845947  215.062439  118.790344    0.590710     26   handbag
2  219.549438  91.001450  245.566772  128.410980    0.546561     28  suitcase
3  190.115112  28.154009  217.299866   87.963417    0.507633      0    person
4  219.012680  90.596786  245.095016  128.418671    0.276464     26   handbag

Saved 1 image to output42

Object counts: name
person      2
handbag     2
suitcase    1
Name: count, dtype: int64


Results for thumbnail_1984.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0   38.970001    8.946289  189.883606  170.122406    0.928110      0  person
1  207.236069   33.061363  319.765259  166.005798    0.692671      0  person
2   10.784012  140.824539   24.866302  157.100479    0.256825     41     cup

Saved 1 image to output43

Object counts: name
person    2
cup       1
Name: count, dtype: int64


Results for thumbnail_0014.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output44

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_1089.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  155.791061   7.510849  301.314453  171.474182    0.726388      0  person
1   63.012779  65.291687  154.914551  168.663681    0.533001     16     dog

Saved 1 image to output45

Object counts: name
person    1
dog       1
Name: count, dtype: int64


Results for thumbnail_0095.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  70.442627  74.804337  126.911118  171.629593    0.252302      0  person

Saved 1 image to output46

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3964.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0    1.730202   2.637009  113.904175  167.680542    0.851113      0  person
1   87.937302  19.548653  228.495743  171.325623    0.835442      0  person
2  221.666534   3.076134  319.228485  171.164429    0.785290      0  person
3  205.645477   0.476009  247.997864   33.175938    0.780001     74   clock

Saved 1 image to output47

Object counts: name
person    3
clock     1
Name: count, dtype: int64


Results for thumbnail_5120.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  212.983566  94.568283  227.233231  124.106827    0.484411      0  person
1  187.584137  88.796165  198.898926  123.093910    0.363110      0  person

Saved 1 image to output48

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_0200.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0  139.897064    0.000000  289.053558  152.858612    0.802378      0  person
1  211.177933  120.213455  256.772552  172.690262    0.668188     41     cup
2    0.000000   82.700356   55.418816  163.130249    0.623903     15     cat

Saved 1 image to output49

Object counts: name
person    1
cup       1
cat       1
Name: count, dtype: int64


Results for thumbnail_5513.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output50

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_0529.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output51

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_2182.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  117.981087   5.748840  177.537598   88.306969    0.607657      0  person
1  169.251160  63.801407  222.469238  169.500259    0.555439      0  person

Saved 1 image to output52

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_4626.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  101.158691  19.309608  172.481598  168.649536    0.846888      0  person
1  160.050735  83.969330  170.573074  106.003815    0.288472     41     cup

Saved 1 image to output53

Object counts: name
person    1
cup       1
Name: count, dtype: int64


Results for thumbnail_3445.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  193.976654    6.651924  318.543304  170.813904    0.923951      0   
1   30.537571   46.784252  116.442978  169.860519    0.857799      0   
2  129.496887  111.624481  167.422363  127.380661    0.560207     45   
3   48.540405  116.708023  247.623566  171.830215    0.452929     60   
4   15.778946  105.227631   38.750134  171.864441    0.295880     56   
5   99.217392  108.180389  127.545090  122.639557    0.273594     45   

           name  
0        person  
1        person  
2          bowl  
3  dining table  
4         chair  
5          bowl  

Saved 1 image to output54

Object counts: name
person          2
bowl            2
dining table    1
chair           1
Name: count, dtype: int64


Results for thumbnail_5247.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  237.878387  29.059677  320.000000  169.417175    0.816215      0  person
1   10.979935   5.765930  239.388382  171.036133    0.598696      0  person

Saved 1 image to output55

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_1699.jpg:
         xmin       ymin        xmax        ymax  confidence  class      name
0  190.774094  11.055222  320.000000   93.422134    0.409655     24  backpack
1    3.926575  19.543854  192.496094  166.544785    0.402526      0    person
2  289.293640   7.737494  319.914246   33.194191    0.304620     29   frisbee

Saved 1 image to output56

Object counts: name
backpack    1
person      1
frisbee     1
Name: count, dtype: int64


Results for thumbnail_3898.jpg:
         xmin       ymin        xmax        ymax  confidence  class      name
0  114.886719  19.059753  149.560120  102.785683    0.804445      0    person
1  159.604385  16.683743  190.343216   88.616684    0.800794      0    person
2   62.854935  98.494354   93.992386  170.672516    0.728787      0    person
3   89.676476  63.079086  131.346680  150.976059    0.683238      0    person
4    0.339666  62.535774   52.461746  142.538086    0.367685      0    person
5    0.333313  65.679573   52.962120  141.874054    0.295822     20  elephant

Saved 1 image to output57

Object counts: name
person      5
elephant    1
Name: count, dtype: int64


Results for thumbnail_4839.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  82.920807  16.141014  285.400238  169.616791    0.728031      0  person

Saved 1 image to output58

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4574.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output59

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_1377.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   73.470001   6.139549  128.507660  169.341125    0.916058      0  person
1   44.041840  31.264500   77.113052  143.782593    0.856360      0  person
2  187.086700  16.840050  318.248352  166.675720    0.250390     13   bench

Saved 1 image to output60

Object counts: name
person    2
bench     1
Name: count, dtype: int64


Results for thumbnail_3862.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  171.826752  75.508713  240.844025  170.424866    0.918988      0  person
1  276.243866  53.097939  319.812408  169.375824    0.867047      0  person
2  113.146454  51.449036  176.015503  171.291748    0.708953      0  person
3   20.138355  21.432114  122.580246  169.726852    0.670452      0  person

Saved 1 image to output61

Object counts: name
person    4
Name: count, dtype: int64


Results for thumbnail_0378.jpg:
         xmin      ymin        xmax        ymax  confidence  class    name
0  194.087830  0.828224  318.320251  126.794373    0.641911      0  person
1    5.028244  0.846451  153.324051  160.507614    0.257506      0  person

Saved 1 image to output62

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_4545.jpg:
         xmin       ymin        xmax        ymax  confidence  class name
0  139.352173  27.057243  317.233337  169.692352    0.265721     62   tv

Saved 1 image to output63

Object counts: name
tv    1
Name: count, dtype: int64


Results for thumbnail_3979.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  190.761841  41.980858  317.221466  170.238678    0.922442      0  person
1   59.576996   7.895622  197.433319  169.984100    0.821896      0  person
2    0.073750  67.026413   20.864586  143.365295    0.252041      0  person

Saved 1 image to output64

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_3282.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  211.477722   4.655518  320.000000  171.839111    0.910799      0  person
1    0.000000  23.410858  146.875397  171.064514    0.879793      0  person
2  104.470978  10.857124  265.247894  171.711426    0.650472      0  person
3   17.181168   0.000000  130.876068  117.310829    0.408740      0  person

Saved 1 image to output65

Object counts: name
person    4
Name: count, dtype: int64


Results for thumbnail_4012.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  211.324768  53.164761  319.998413  170.327530    0.895948      0  person
1    0.628002   4.830811   80.950760  171.659821    0.845913      0  person
2   53.243706  23.945839  230.730347  170.081024    0.835954      0  person

Saved 1 image to output66

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_1021.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  149.861023  16.046036  291.207458  170.548782    0.921506      0  person
1    1.531708   1.513390  153.490204  171.160492    0.667511      0  person

Saved 1 image to output67

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_1598.jpg:
         xmin       ymin        xmax        ymax  confidence  class  \
0  126.216827  35.160980  211.332458  169.978455    0.783211      0   
1  292.956787  71.623627  319.963989  170.862137    0.396760     67   
2    0.790085   6.666939   87.218994  170.112122    0.333143      1   
3  165.729187  82.807693  176.400269  109.252792    0.254341     27   

         name  
0      person  
1  cell phone  
2     bicycle  
3         tie  

Saved 1 image to output68

Object counts: name
person        1
cell phone    1
bicycle       1
tie           1
Name: count, dtype: int64


Results for thumbnail_5484.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output69

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_2531.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0  170.470856  125.422737  194.173828  139.322113    0.653060     49  orange
1    0.626236    7.589973  176.629318  168.125641    0.546856      0  person
2  134.373108  103.405441  177.702301  133.711777    0.276823     47   apple
3  143.306168  130.781815  169.047806  139.649734    0.276766     47   apple

Saved 1 image to output70

Object counts: name
apple     2
orange    1
person    1
Name: count, dtype: int64


Results for thumbnail_3759.jpg:
         xmin      ymin       xmax        ymax  confidence  class    name
0  185.313995  1.310852  319.36377  170.152939    0.526603      0  person

Saved 1 image to output71

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_5026.jpg:
         xmin        ymin        xmax        ymax  confidence  class   name
0    0.142338   17.425537   67.875641  129.616791    0.938958     54  donut
1  143.749176   53.415813  283.605408  171.608902    0.937688     54  donut
2   64.103645    0.523243  170.689819   76.299057    0.934066     54  donut
3    5.105659  123.523270  149.794250  171.094986    0.886766     54  donut
4  256.606140    0.425198  319.771362   89.695084    0.861960     54  donut
5  289.287476   98.720673  319.910400  170.751129    0.814197     54  donut
6    0.674068    0.000000   89.452621   20.903061    0.691111     54  donut
7  159.298264    0.000000  262.439148   29.352699    0.572505     54  donut

Saved 1 image to output72

Object counts: name
donut    8
Name: count, dtype: int64


Results for thumbnail_0254.jpg:
         xmin       ymin        xmax        ymax  confidence  class  \
0  164.427383  67.048492  273.870483  170.180237    0.913898      0   
1   70.134346  19.490311  205.201080  170.812210    0.854274      0   
2    5.394274   4.411087   66.836456  169.848083    0.328622     72   

           name  
0        person  
1        person  
2  refrigerator  

Saved 1 image to output73

Object counts: name
person          2
refrigerator    1
Name: count, dtype: int64


Results for thumbnail_4042.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  118.218582  26.487823  275.852234  170.865799    0.909281      0  person
1  237.848770  98.459358  268.461914  125.023445    0.393558      0  person

Saved 1 image to output74

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_0428.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output75

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_5444.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output76

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_0769.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  183.215271   5.020081  262.348328  171.989487    0.909600      0  person
1   80.604202  30.882828  156.781235  171.124176    0.868988      0  person

Saved 1 image to output77

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_3731.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  41.450253  23.539391  140.269623  170.491852    0.852279      0  person

Saved 1 image to output78

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3078.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output79

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_1116.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  229.355499    4.368484  310.134003  169.587250    0.918370      0   
1   80.349701   66.505928  138.206833  171.436829    0.774849      0   
2  105.281227  102.943939  166.011078  172.149261    0.382820     26   
3  191.874115   63.627968  232.225433  136.065613    0.344178     58   
4   81.738907   91.346222  134.283218  171.333679    0.258839     26   

           name  
0        person  
1        person  
2       handbag  
3  potted plant  
4       handbag  

Saved 1 image to output80

Object counts: name
person          2
handbag         2
potted plant    1
Name: count, dtype: int64


Results for thumbnail_2232.jpg:
         xmin        ymin        xmax        ymax  confidence  class     name
0  143.155670   26.374390  201.231689  171.148529    0.826259      0   person
1  206.280121   44.986691  264.959686  169.473175    0.787986      0   person
2  166.232986  101.004951  206.110367  140.483139    0.414332     26  handbag
3   79.511681   45.549789   95.488655   79.918976    0.399651      0   person
4    0.874062    1.985268  145.176880  167.816589    0.277828      6    train

Saved 1 image to output81

Object counts: name
person     3
handbag    1
train      1
Name: count, dtype: int64


Results for thumbnail_0721.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  101.000114    8.723999  308.137543  169.073715    0.681804      0   
1  166.830383  110.730492  203.220947  142.293533    0.625566     65   
2  206.456589   62.827759  234.059677   93.706116    0.412990     67   

         name  
0      person  
1      remote  
2  cell phone  

Saved 1 image to output82

Object counts: name
person        1
remote        1
cell phone    1
Name: count, dtype: int64


Results for thumbnail_1710.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output83

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_0364.jpg:
        xmin       ymin        xmax        ymax  confidence  class    name
0  36.987572   0.463600  233.869476  168.912323    0.495480      0  person
1   0.125710  42.961285   75.019386  169.528183    0.477114      0  person

Saved 1 image to output84

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_4016.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  208.082336  54.720047  315.362915  170.697571    0.897045      0  person
1   59.007256  11.698898  217.114288  168.307190    0.863105      0  person
2    0.637535   4.217789   83.680183  170.430283    0.847649      0  person

Saved 1 image to output85

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_3806.jpg:
         xmin        ymin        xmax        ymax  confidence  class  \
0  188.709305   28.699776  271.926575  169.180725    0.890608      0   
1   80.782745  119.574829   99.778595  147.630920    0.572172     58   
2  223.176270   61.173210  231.308838   72.060478    0.298346     67   
3  103.744308  136.159195  129.783234  167.233261    0.297070     58   
4  107.466209   68.414505  122.635109  101.645798    0.276891     58   
5   62.756264  129.628647   81.561714  169.486740    0.270850     58   
6   93.042259   87.998512  108.758293  114.643257    0.265385     58   
7   78.901833   53.107529  106.170174   91.836105    0.263454     58   

           name  
0        person  
1  potted plant  
2    cell phone  
3  potted plant  
4  potted plant  
5  potted plant  
6  potted plant  
7  potted plant  

Saved 1 image to output86

Object counts: name
potted plant    6
person          1
cell phone      1
Name: count, dtype: int64


Results for thumbnail_1445.jpg:
         xmin     ymin        xmax        ymax  confidence  class    name
0  122.655396  3.49102  279.600311  169.836914    0.947074      0  person

Saved 1 image to output87

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4095.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  221.555740  71.861191  264.665710  171.132767    0.911868      0  person
1  156.829025  72.107063  204.332443  169.934998    0.872741      0  person
2   61.355408  81.234375   78.703644  133.764038    0.672436      0  person

Saved 1 image to output88

Object counts: name
person    3
Name: count, dtype: int64


Results for thumbnail_0875.jpg:
        xmin       ymin        xmax        ymax  confidence  class  \
0  96.757240   9.862465  233.891357  171.548645    0.916444      0   
1  73.371826   2.207262  109.689484   33.727257    0.605982     58   
2   0.000000   6.410263   73.038467  141.025604    0.471249     56   
3  39.984402  21.511345  132.989624  153.620514    0.379645     56   
4   0.915382  71.377609   76.028465  154.904160    0.275209     60   
5   6.703049  70.831055   78.062180  154.202209    0.252643     56   

           name  
0        person  
1  potted plant  
2         chair  
3         chair  
4  dining table  
5         chair  

Saved 1 image to output89

Object counts: name
chair           3
person          1
potted plant    1
dining table    1
Name: count, dtype: int64


Results for thumbnail_0247.jpg:
         xmin       ymin        xmax       ymax  confidence  class    name
0  112.381332  22.742783  217.997726  172.18782    0.887781      0  person

Saved 1 image to output90

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4053.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0   70.264435   5.718681  184.477112  169.637390    0.899587      0  person
1  165.884216  55.873562  278.725311  170.362183    0.787739      0  person

Saved 1 image to output91

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_5119.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  210.857208  94.237427  220.066742  124.031525     0.61233      0  person

Saved 1 image to output92

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3992.jpg:
        xmin      ymin        xmax        ymax  confidence  class    name
0  97.227501  4.233742  279.014252  171.454285     0.58972      0  person

Saved 1 image to output93

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3569.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  173.987579  35.839203  312.622864  172.396912    0.911581      0  person
1    0.000000   1.292839  154.701416  171.674835    0.905045      0  person

Saved 1 image to output94

Object counts: name
person    2
Name: count, dtype: int64


Results for thumbnail_5238.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  203.328339  17.522545  289.887665  171.215302    0.931066      0  person

Saved 1 image to output95

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_3768.jpg:
         xmin       ymin        xmax       ymax  confidence  class  name
0  178.414078  61.598492  271.140656  170.68692    0.320065     14  bird

Saved 1 image to output96

Object counts: name
bird    1
Name: count, dtype: int64


Results for thumbnail_2453.jpg:
         xmin        ymin        xmax        ymax  confidence  class    name
0   89.601028    2.355194  273.311035  168.651642    0.953322      0  person
1  158.248856  104.262451  197.786118  171.557343    0.907252     27     tie

Saved 1 image to output97

Object counts: name
person    1
tie       1
Name: count, dtype: int64


Results for thumbnail_4515.jpg:
Empty DataFrame
Columns: [xmin, ymin, xmax, ymax, confidence, class, name]
Index: []

Saved 1 image to output98

Object counts: Series([], Name: count, dtype: int64)


Results for thumbnail_4629.jpg:
         xmin       ymin        xmax        ymax  confidence  class    name
0  105.686798  18.330505  197.794342  170.699188    0.844963      0  person

Saved 1 image to output99

Object counts: name
person    1
Name: count, dtype: int64


Results for thumbnail_4249.jpg:
         xmin       ymin   xmax        ymax  confidence  class    name
0  119.076599  21.926826  320.0  171.252808    0.487322      0  person

Saved 1 image to output100

Object counts: name
person    1
Name: count, dtype: int64

print("Total counts of detected objects:")
for object_type, count in total_counts.items():
    print(f"{object_type}: {count}")

Total counts of detected objects:
4: 3
2: 31
1: 85
3: 11
5: 2
8: 1
6: 1

---

SUMMARY

Film Description

Honey, I Shrunk the Kids

1989 PG 1h 33m

From the IMDB website:

"The scientist father of a teenage girl and boy accidentally shrinks his and two other neighborhood teens to the size of insects. Now the teens must fight diminutive dangers as the father searches for them."

IMDB website. (n.d.). imdb.com. Retrieved April 27, 2024, from https://www.imdb.com/title/tt0097523/

In 'Honey, I Shrunk the Kids,' an eccentric inventor, Wayne Szalinski, accidentally shrinks his and his neighbor's children with his experimental shrink ray. The miniature kids must navigate a perilous journey across their now-gigantic backyard, encountering obstacles like insects and sprinklers, as they try to return home.

The film is notable for its creative visual effects that magnify ordinary environments into epic landscapes. It's a blend of adventure, humor, and family dynamics, ultimately showcasing the children's resourcefulness and the parents' determination to rescue their kids. The movie was a commercial success and spawned a franchise including sequels and a television series.

Methods Summary

This section should highlight methods you used in your exploratory analysis. You should include at least one clustering technique or develop another way to relate frames to other frames. You should also consider dimensionality reduction.

Each thumbnail is one frame of the Movie. Each thumbnail has been analyzed and embedded via CLIP. Referenced Model: https://replicate.com/andreasjansson/clip-features/examples .

For hypothesis 1 we used the following methods

1. Explore CLIP through a natural query for objects and investigate the surounding frames.
2. Look at the Euclidean distance of similar frames.
3. Explore similar frames for insights.

Hunches and Hypotheses

This section should summarize the questions that you asked about the film that could potentially be answered by exploratory analysis. You should ask at least three questions.

Hypothesis 1

• Using Clip to find an object and observe ~5 frames before and after and interpret the results.
• A hypothesis we discussed was that certain characters, or objects in the movie would be used for dramatic or comedic effect. We are using CLIP to identify these objects and then explore the surrounding frames to discover any evidence to support or refute this hypothesis.

Hypothesis 2

• Scene Consistency and Transition - Frames that are visually and thematically similar cluster together tightly in t-SNE and PCA visualizations, and distinct clusters correspond to different scenes or settings in the movie.
• Rationale: This hypothesis tests the ability of CLIP embeddings, which capture both visual and semantic content, to differentiate between distinct scenes based on their visual content and thematic elements.

Hypothesis 3

• Look for a model that can analyze an image to identify the type of objects in the image
• Use counter to count the number of each object in the image

Results and Interpretation

Hypothesis 1:

• Begin by looking for the top images of a dog, lawnmower and an ant returned from CLIP queries.
• Explore the frames surrounding the images identified by the above queries.

These are the top matches returned by CLIP.

1. 'a photo of a white and brown dog' = thumbnail_5039.jpg
2. 'a photo of a lawnmower' = thumbnail_4734.jpg
3. 'a photo of an ant' = thumbnail_3231.jpg

The indexes we are interested in are 5038, 4733 and 3230.

• For the dog image:

From this sequence of frames, we see the children running, the parents talking, the dog looking up at the table, the kids waving their arms for attention and ends with the father narrowly avoiding accidentally eating his shrunken son. The dog being in the center anchors the dramatic tension of the scene. He almost looks like he is trying to warn the father or somehow knows more than the humans. Sentient animals with superior knowledge or wisdom than humans has been a trope used in several other films which depict animals (especially in comedies) as central figures.

• For the lawnmower image:

In this scene a neighbor kid is remotely controlling the lawnmower while the children are in the grass. This is a clear action scene with a threat elliciting dramatic flight.

There doesn't seem to be many similar frames in the movie for the lawnmower. This is probably because it is a singular threat evnt used for dramatic tension in a single scene. It does look like it may appear around 790, so let's see if that is foreshadowing.

In the scene around index 790 we see the son showing the remote controlled lawnmower to the boy who appears later controlling the mower in the flight scene. So this is an example of foreshadowing in a film that we identified by using the data to point us to something interesting!

• For the ant image:

In this scene, we see a boy easting some white substance, followed by a girl who appears scared or surprised. The next frame shows the cause, which is an ant menacing the kids, which is noticed by the boy as well, and the kids seem to begin fleeing the threat.

The ant seems to be a later threat in the film, possibly also foreshadowed around 2180. However it appears to show up for an extended time around 4400.

2180 looks like it matches the ant scene because of the setting. They both have extensive vegetation in the scene, which probably accounts for the lower distance than the later match around 4400.

This identified the same scene that we discovered through exploration.

• We began with a hypothesis that the dog would be comedic relief and the ant and lawnmower would be for dramatic effect or threat of danger.We used CLIP to identify the inital frames of interest, then proceeded analyzing the target frames using the Euclidian distance, performing more exploration of similar frames and then circling back to a more specific CLIP question. The original hypothesis that these objects would be used as comedic relief or dramatic seems to have some justification, as the lawnmower and ant appear intially as dramatic threats, but from this discovery, it looks like the ant actually fought a scorpion and saved the kids, which indicated it was also used to elicit a sympathetic response from the audience.

Hypothesis 2:

• After removing much of the noise from opening and closing credits frames and applying t-SNE and K-means we examined samples from identified clusters and we can observe that frames correspond to specific scenes or types of scenes (indoor vs outdoor, calm vs action-packed). Embeddings effectively capture scene-specific features and can be used to segment the movie based on visual content. Also, clusters seem to reflect the proper timeframe and transition of the movie. t-SNE seemed to work better than PCA at handling and splitting clusters with visual correlation.

Hypothesis 3:

• We used Yolo model, an open source model we got from the internet, to identify objects in an image. We initially analyzed the entire images in the thumpnail_folder, but found out that it wasn't a good idea because the code was slow, so we decided to analyze the first 100 images in the image frame.The detected results are stored in a DataFrame df, which is printed out. The DataFrame contains columns such as xmin, xmax, ymin, ymax, confidence, and name of the detected object. The detected objects are overlaid on the image with bounding boxes and the image is displayed using matplotlib. After all images are processed, the total counts of detected objects across all images are printed. This provides a summary of how many of each type of object were detected in the analyzed images.

Reflection

Reflect on your process of analysis. What worked well and did not work well? Describe the limitations of the work and describe what you would work on with more time.

We initially began exploring CLIP to discover the capabilities it provided. We were impressed with the model's ability to return results from simple english questions, so we brainstormed for some ideas that we could explore it as well as other tools to compare and contrast CLIP to other tools and also to discover what it could tell us about the film itself.

We came up with the hypothesis that we could identify some objects in the film and explore whether they provided an assumed purpose in the movie. We did achieve (subjectively) a confirmation tthat the objects were used in a way that coincided with our hypothesis. What was interesting is the way that our initial exploration led to a secondary use that effectively used CLIP to identify a further area of research. This resulted in the surprising find that the ant seems to not only be used as a threat but also as an ally to evoke both fear and empathy.